The most common question we get from people considering SoloScan is some version of this: "I already use ChatGPT. Why would I need anything else?"
It is a fair question. ChatGPT is a great product. For most knowledge work, it is the right tool. This article is not about why ChatGPT is bad. It is about the specific kinds of work where ChatGPT is not the right answer, and where SoloScan exists to fill that gap.
If you read this and conclude ChatGPT is enough for what you do, that is a fine answer. Use ChatGPT. We will not be hurt.
What ChatGPT is great at
A short list of work where ChatGPT (and Claude, Gemini, Copilot) is hard to beat:
- Writing and editing prose.
- Brainstorming, outlining, and structuring ideas.
- Summarizing publicly available content.
- Explaining technical concepts.
- Drafting emails, social posts, and marketing copy.
- Code generation and debugging.
- Answering questions about general knowledge.
For all of these, cloud AI is faster, more capable, and almost always the right pick. The current frontier models (GPT-4 class and above) are noticeably stronger at deep reasoning than any model you can run on a laptop today. If your work falls in this list, stop reading and go use ChatGPT.
Where ChatGPT stops being the answer
There is one category of work where cloud AI is the wrong tool: anything involving someone else's confidential information.
If you are an accountant reviewing a client's payroll ledger, an auditor working through a company's annual report, a lawyer reading a draft contract, a therapist transcribing session notes, or a consultant analyzing a private board deck — the documents you handle do not belong to you. They belong to the people who hired you. And those people did not agree to have their information sent to OpenAI, Microsoft, Anthropic, or Google.
This is not paranoia. It is professional obligation. Three things make uploading confidential client documents to ChatGPT a real problem:
-
Your contract with the client. Most professional engagements include a confidentiality clause that prohibits sharing the client's data with third parties. ChatGPT is a third party.
-
Your professional body's rules. Bar associations, accountancy boards, medical boards, and similar regulators have explicit rules about confidentiality. Most have not yet caught up with AI specifically, but the general rule (do not share client information without consent) covers it.
-
Data protection law. GDPR in the EU, similar regimes in other jurisdictions, sometimes require explicit data processor agreements before sharing personal information with a service like OpenAI. The agreements exist but are rarely signed individually by professionals.
The legal team at OpenAI will tell you their enterprise tier handles this. They are partly right. But the enterprise tier costs significantly more than the consumer one, and even with the right agreements in place, you have moved the data outside your direct control. For some clients and some jurisdictions, that is the violation, regardless of what the receiving company promises to do with it.
What SoloScan does differently
SoloScan runs the AI model on your own computer. The document never leaves your machine. There is no cloud upload, no third-party server, no data processor agreement to sign.
Concretely:
- You open SoloScan and drop in a PDF, DOCX, or scanned image.
- A local AI model (we use Llama 3.2 via Ollama, both running on your computer) reads the document.
- You ask questions. The model answers, with the document text as context.
- The conversation is saved to a local SQLite file in your user folder. No server involved.
The trade-off is real and we are honest about it:
- The local model is not as smart as GPT-4 or Claude. For deep multi-document reasoning, ChatGPT will be better.
- The setup requires installing Ollama once (about ten minutes). ChatGPT is one click away in a browser.
- SoloScan is paid software ($39/month). ChatGPT has a free tier.
For most knowledge work, those trade-offs make ChatGPT the right choice. For confidential document work, they are worth it because the alternative is not "use ChatGPT instead" — it is "do this work entirely by hand, without AI help at all".
A side-by-side, task by task
Here is a more concrete comparison, broken down by what you might actually try to do.
Reading a 50-page contract
ChatGPT (Pro): Strong. Will summarize, find clauses, flag risks. But uploads the contract to OpenAI's servers. Not appropriate for client contracts under most engagement agreements.
SoloScan: Strong. Will summarize, find clauses, flag risks. Stays on your machine. Quality is slightly lower than GPT-4 on subtle legal reasoning, but more than sufficient for first-pass review.
Summarizing a public earnings report
ChatGPT: Better. Public document, no confidentiality concerns, GPT-4 is sharper.
SoloScan: Works fine. But you do not need the privacy guarantee for public data, and ChatGPT will give you a cleaner summary.
Extracting line items and VAT totals from client receipts
ChatGPT: Will do it. Sends client expense data to a third party.
SoloScan: Will do it. Stays on your machine. This is the canonical use case.
Drafting a client-facing report based on private analysis
ChatGPT: Useful for the writing itself, but you should not paste the private analysis in as context.
SoloScan: Useful for both. Extract findings from the source documents with SoloScan, then optionally use ChatGPT only on the final, anonymized prose.
Casual general-knowledge questions
ChatGPT: Obviously the right tool.
SoloScan: Not designed for this. SoloScan is document-focused.
The honest summary
ChatGPT is the right tool for most of your work. SoloScan is the right tool for the slice of your work where confidentiality is not optional.
Many professionals will end up using both. That is a perfectly reasonable arrangement. Use ChatGPT for the public, generic, and drafting work. Use SoloScan for the documents that have a real client behind them.
If you want to try the latter, SoloScan is free to download and analyzes the first 1,000 characters of any document on the free plan. Drop in a document that you would not be comfortable uploading to ChatGPT, and see whether the workflow fits.